Modelling Cryogenic Hydrogen Jet Dispersion: CFD-POD-ML Insights

Abstract: Safe and efficient methods for storing and transporting liquid hydrogen are crucial for decarbonisation. This study develops a precise method for analysing complex cryogenic hydrogen releases, integrating large eddy simulation (LES), proper orthogonal decomposition (POD), and a novel hybrid machine learning model. Focusing on LES predictions of Sandia's experiment on cryogenic hydrogen jet release at stagnation pressure (5 bar) and temperature (50 K), POD extracts dominant modes, demonstrating that the first 10 modes effectively predict hydrogen dispersion. The hybrid model combines POD with Long Short-Term Memory neural networks and Simple Exponential Smoothing techniques to forecast temporal coefficients, showing acceptable agreement with LES and POD predictions. This research provides significant insights into effective approaches in cryogenic hydrogen applications, significantly reducing the time and costs in numerical and experimental investigations. Keywords: Liquid Cryogenic Hydrogen, Proper Orthogonal Decomposition, Long Short-Term Memory, Simple Exponential Smoothing, Large Eddy Simulation